xref: /openbmc/linux/fs/btrfs/raid56.c (revision 6d99a79c)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2012 Fusion-io  All rights reserved.
4  * Copyright (C) 2012 Intel Corp. All rights reserved.
5  */
6 
7 #include <linux/sched.h>
8 #include <linux/bio.h>
9 #include <linux/slab.h>
10 #include <linux/blkdev.h>
11 #include <linux/raid/pq.h>
12 #include <linux/hash.h>
13 #include <linux/list_sort.h>
14 #include <linux/raid/xor.h>
15 #include <linux/mm.h>
16 #include "ctree.h"
17 #include "disk-io.h"
18 #include "volumes.h"
19 #include "raid56.h"
20 #include "async-thread.h"
21 
22 /* set when additional merges to this rbio are not allowed */
23 #define RBIO_RMW_LOCKED_BIT	1
24 
25 /*
26  * set when this rbio is sitting in the hash, but it is just a cache
27  * of past RMW
28  */
29 #define RBIO_CACHE_BIT		2
30 
31 /*
32  * set when it is safe to trust the stripe_pages for caching
33  */
34 #define RBIO_CACHE_READY_BIT	3
35 
36 #define RBIO_CACHE_SIZE 1024
37 
38 enum btrfs_rbio_ops {
39 	BTRFS_RBIO_WRITE,
40 	BTRFS_RBIO_READ_REBUILD,
41 	BTRFS_RBIO_PARITY_SCRUB,
42 	BTRFS_RBIO_REBUILD_MISSING,
43 };
44 
45 struct btrfs_raid_bio {
46 	struct btrfs_fs_info *fs_info;
47 	struct btrfs_bio *bbio;
48 
49 	/* while we're doing rmw on a stripe
50 	 * we put it into a hash table so we can
51 	 * lock the stripe and merge more rbios
52 	 * into it.
53 	 */
54 	struct list_head hash_list;
55 
56 	/*
57 	 * LRU list for the stripe cache
58 	 */
59 	struct list_head stripe_cache;
60 
61 	/*
62 	 * for scheduling work in the helper threads
63 	 */
64 	struct btrfs_work work;
65 
66 	/*
67 	 * bio list and bio_list_lock are used
68 	 * to add more bios into the stripe
69 	 * in hopes of avoiding the full rmw
70 	 */
71 	struct bio_list bio_list;
72 	spinlock_t bio_list_lock;
73 
74 	/* also protected by the bio_list_lock, the
75 	 * plug list is used by the plugging code
76 	 * to collect partial bios while plugged.  The
77 	 * stripe locking code also uses it to hand off
78 	 * the stripe lock to the next pending IO
79 	 */
80 	struct list_head plug_list;
81 
82 	/*
83 	 * flags that tell us if it is safe to
84 	 * merge with this bio
85 	 */
86 	unsigned long flags;
87 
88 	/* size of each individual stripe on disk */
89 	int stripe_len;
90 
91 	/* number of data stripes (no p/q) */
92 	int nr_data;
93 
94 	int real_stripes;
95 
96 	int stripe_npages;
97 	/*
98 	 * set if we're doing a parity rebuild
99 	 * for a read from higher up, which is handled
100 	 * differently from a parity rebuild as part of
101 	 * rmw
102 	 */
103 	enum btrfs_rbio_ops operation;
104 
105 	/* first bad stripe */
106 	int faila;
107 
108 	/* second bad stripe (for raid6 use) */
109 	int failb;
110 
111 	int scrubp;
112 	/*
113 	 * number of pages needed to represent the full
114 	 * stripe
115 	 */
116 	int nr_pages;
117 
118 	/*
119 	 * size of all the bios in the bio_list.  This
120 	 * helps us decide if the rbio maps to a full
121 	 * stripe or not
122 	 */
123 	int bio_list_bytes;
124 
125 	int generic_bio_cnt;
126 
127 	refcount_t refs;
128 
129 	atomic_t stripes_pending;
130 
131 	atomic_t error;
132 	/*
133 	 * these are two arrays of pointers.  We allocate the
134 	 * rbio big enough to hold them both and setup their
135 	 * locations when the rbio is allocated
136 	 */
137 
138 	/* pointers to pages that we allocated for
139 	 * reading/writing stripes directly from the disk (including P/Q)
140 	 */
141 	struct page **stripe_pages;
142 
143 	/*
144 	 * pointers to the pages in the bio_list.  Stored
145 	 * here for faster lookup
146 	 */
147 	struct page **bio_pages;
148 
149 	/*
150 	 * bitmap to record which horizontal stripe has data
151 	 */
152 	unsigned long *dbitmap;
153 
154 	/* allocated with real_stripes-many pointers for finish_*() calls */
155 	void **finish_pointers;
156 
157 	/* allocated with stripe_npages-many bits for finish_*() calls */
158 	unsigned long *finish_pbitmap;
159 };
160 
161 static int __raid56_parity_recover(struct btrfs_raid_bio *rbio);
162 static noinline void finish_rmw(struct btrfs_raid_bio *rbio);
163 static void rmw_work(struct btrfs_work *work);
164 static void read_rebuild_work(struct btrfs_work *work);
165 static int fail_bio_stripe(struct btrfs_raid_bio *rbio, struct bio *bio);
166 static int fail_rbio_index(struct btrfs_raid_bio *rbio, int failed);
167 static void __free_raid_bio(struct btrfs_raid_bio *rbio);
168 static void index_rbio_pages(struct btrfs_raid_bio *rbio);
169 static int alloc_rbio_pages(struct btrfs_raid_bio *rbio);
170 
171 static noinline void finish_parity_scrub(struct btrfs_raid_bio *rbio,
172 					 int need_check);
173 static void scrub_parity_work(struct btrfs_work *work);
174 
175 static void start_async_work(struct btrfs_raid_bio *rbio, btrfs_func_t work_func)
176 {
177 	btrfs_init_work(&rbio->work, btrfs_rmw_helper, work_func, NULL, NULL);
178 	btrfs_queue_work(rbio->fs_info->rmw_workers, &rbio->work);
179 }
180 
181 /*
182  * the stripe hash table is used for locking, and to collect
183  * bios in hopes of making a full stripe
184  */
185 int btrfs_alloc_stripe_hash_table(struct btrfs_fs_info *info)
186 {
187 	struct btrfs_stripe_hash_table *table;
188 	struct btrfs_stripe_hash_table *x;
189 	struct btrfs_stripe_hash *cur;
190 	struct btrfs_stripe_hash *h;
191 	int num_entries = 1 << BTRFS_STRIPE_HASH_TABLE_BITS;
192 	int i;
193 	int table_size;
194 
195 	if (info->stripe_hash_table)
196 		return 0;
197 
198 	/*
199 	 * The table is large, starting with order 4 and can go as high as
200 	 * order 7 in case lock debugging is turned on.
201 	 *
202 	 * Try harder to allocate and fallback to vmalloc to lower the chance
203 	 * of a failing mount.
204 	 */
205 	table_size = sizeof(*table) + sizeof(*h) * num_entries;
206 	table = kvzalloc(table_size, GFP_KERNEL);
207 	if (!table)
208 		return -ENOMEM;
209 
210 	spin_lock_init(&table->cache_lock);
211 	INIT_LIST_HEAD(&table->stripe_cache);
212 
213 	h = table->table;
214 
215 	for (i = 0; i < num_entries; i++) {
216 		cur = h + i;
217 		INIT_LIST_HEAD(&cur->hash_list);
218 		spin_lock_init(&cur->lock);
219 	}
220 
221 	x = cmpxchg(&info->stripe_hash_table, NULL, table);
222 	if (x)
223 		kvfree(x);
224 	return 0;
225 }
226 
227 /*
228  * caching an rbio means to copy anything from the
229  * bio_pages array into the stripe_pages array.  We
230  * use the page uptodate bit in the stripe cache array
231  * to indicate if it has valid data
232  *
233  * once the caching is done, we set the cache ready
234  * bit.
235  */
236 static void cache_rbio_pages(struct btrfs_raid_bio *rbio)
237 {
238 	int i;
239 	char *s;
240 	char *d;
241 	int ret;
242 
243 	ret = alloc_rbio_pages(rbio);
244 	if (ret)
245 		return;
246 
247 	for (i = 0; i < rbio->nr_pages; i++) {
248 		if (!rbio->bio_pages[i])
249 			continue;
250 
251 		s = kmap(rbio->bio_pages[i]);
252 		d = kmap(rbio->stripe_pages[i]);
253 
254 		copy_page(d, s);
255 
256 		kunmap(rbio->bio_pages[i]);
257 		kunmap(rbio->stripe_pages[i]);
258 		SetPageUptodate(rbio->stripe_pages[i]);
259 	}
260 	set_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
261 }
262 
263 /*
264  * we hash on the first logical address of the stripe
265  */
266 static int rbio_bucket(struct btrfs_raid_bio *rbio)
267 {
268 	u64 num = rbio->bbio->raid_map[0];
269 
270 	/*
271 	 * we shift down quite a bit.  We're using byte
272 	 * addressing, and most of the lower bits are zeros.
273 	 * This tends to upset hash_64, and it consistently
274 	 * returns just one or two different values.
275 	 *
276 	 * shifting off the lower bits fixes things.
277 	 */
278 	return hash_64(num >> 16, BTRFS_STRIPE_HASH_TABLE_BITS);
279 }
280 
281 /*
282  * stealing an rbio means taking all the uptodate pages from the stripe
283  * array in the source rbio and putting them into the destination rbio
284  */
285 static void steal_rbio(struct btrfs_raid_bio *src, struct btrfs_raid_bio *dest)
286 {
287 	int i;
288 	struct page *s;
289 	struct page *d;
290 
291 	if (!test_bit(RBIO_CACHE_READY_BIT, &src->flags))
292 		return;
293 
294 	for (i = 0; i < dest->nr_pages; i++) {
295 		s = src->stripe_pages[i];
296 		if (!s || !PageUptodate(s)) {
297 			continue;
298 		}
299 
300 		d = dest->stripe_pages[i];
301 		if (d)
302 			__free_page(d);
303 
304 		dest->stripe_pages[i] = s;
305 		src->stripe_pages[i] = NULL;
306 	}
307 }
308 
309 /*
310  * merging means we take the bio_list from the victim and
311  * splice it into the destination.  The victim should
312  * be discarded afterwards.
313  *
314  * must be called with dest->rbio_list_lock held
315  */
316 static void merge_rbio(struct btrfs_raid_bio *dest,
317 		       struct btrfs_raid_bio *victim)
318 {
319 	bio_list_merge(&dest->bio_list, &victim->bio_list);
320 	dest->bio_list_bytes += victim->bio_list_bytes;
321 	dest->generic_bio_cnt += victim->generic_bio_cnt;
322 	bio_list_init(&victim->bio_list);
323 }
324 
325 /*
326  * used to prune items that are in the cache.  The caller
327  * must hold the hash table lock.
328  */
329 static void __remove_rbio_from_cache(struct btrfs_raid_bio *rbio)
330 {
331 	int bucket = rbio_bucket(rbio);
332 	struct btrfs_stripe_hash_table *table;
333 	struct btrfs_stripe_hash *h;
334 	int freeit = 0;
335 
336 	/*
337 	 * check the bit again under the hash table lock.
338 	 */
339 	if (!test_bit(RBIO_CACHE_BIT, &rbio->flags))
340 		return;
341 
342 	table = rbio->fs_info->stripe_hash_table;
343 	h = table->table + bucket;
344 
345 	/* hold the lock for the bucket because we may be
346 	 * removing it from the hash table
347 	 */
348 	spin_lock(&h->lock);
349 
350 	/*
351 	 * hold the lock for the bio list because we need
352 	 * to make sure the bio list is empty
353 	 */
354 	spin_lock(&rbio->bio_list_lock);
355 
356 	if (test_and_clear_bit(RBIO_CACHE_BIT, &rbio->flags)) {
357 		list_del_init(&rbio->stripe_cache);
358 		table->cache_size -= 1;
359 		freeit = 1;
360 
361 		/* if the bio list isn't empty, this rbio is
362 		 * still involved in an IO.  We take it out
363 		 * of the cache list, and drop the ref that
364 		 * was held for the list.
365 		 *
366 		 * If the bio_list was empty, we also remove
367 		 * the rbio from the hash_table, and drop
368 		 * the corresponding ref
369 		 */
370 		if (bio_list_empty(&rbio->bio_list)) {
371 			if (!list_empty(&rbio->hash_list)) {
372 				list_del_init(&rbio->hash_list);
373 				refcount_dec(&rbio->refs);
374 				BUG_ON(!list_empty(&rbio->plug_list));
375 			}
376 		}
377 	}
378 
379 	spin_unlock(&rbio->bio_list_lock);
380 	spin_unlock(&h->lock);
381 
382 	if (freeit)
383 		__free_raid_bio(rbio);
384 }
385 
386 /*
387  * prune a given rbio from the cache
388  */
389 static void remove_rbio_from_cache(struct btrfs_raid_bio *rbio)
390 {
391 	struct btrfs_stripe_hash_table *table;
392 	unsigned long flags;
393 
394 	if (!test_bit(RBIO_CACHE_BIT, &rbio->flags))
395 		return;
396 
397 	table = rbio->fs_info->stripe_hash_table;
398 
399 	spin_lock_irqsave(&table->cache_lock, flags);
400 	__remove_rbio_from_cache(rbio);
401 	spin_unlock_irqrestore(&table->cache_lock, flags);
402 }
403 
404 /*
405  * remove everything in the cache
406  */
407 static void btrfs_clear_rbio_cache(struct btrfs_fs_info *info)
408 {
409 	struct btrfs_stripe_hash_table *table;
410 	unsigned long flags;
411 	struct btrfs_raid_bio *rbio;
412 
413 	table = info->stripe_hash_table;
414 
415 	spin_lock_irqsave(&table->cache_lock, flags);
416 	while (!list_empty(&table->stripe_cache)) {
417 		rbio = list_entry(table->stripe_cache.next,
418 				  struct btrfs_raid_bio,
419 				  stripe_cache);
420 		__remove_rbio_from_cache(rbio);
421 	}
422 	spin_unlock_irqrestore(&table->cache_lock, flags);
423 }
424 
425 /*
426  * remove all cached entries and free the hash table
427  * used by unmount
428  */
429 void btrfs_free_stripe_hash_table(struct btrfs_fs_info *info)
430 {
431 	if (!info->stripe_hash_table)
432 		return;
433 	btrfs_clear_rbio_cache(info);
434 	kvfree(info->stripe_hash_table);
435 	info->stripe_hash_table = NULL;
436 }
437 
438 /*
439  * insert an rbio into the stripe cache.  It
440  * must have already been prepared by calling
441  * cache_rbio_pages
442  *
443  * If this rbio was already cached, it gets
444  * moved to the front of the lru.
445  *
446  * If the size of the rbio cache is too big, we
447  * prune an item.
448  */
449 static void cache_rbio(struct btrfs_raid_bio *rbio)
450 {
451 	struct btrfs_stripe_hash_table *table;
452 	unsigned long flags;
453 
454 	if (!test_bit(RBIO_CACHE_READY_BIT, &rbio->flags))
455 		return;
456 
457 	table = rbio->fs_info->stripe_hash_table;
458 
459 	spin_lock_irqsave(&table->cache_lock, flags);
460 	spin_lock(&rbio->bio_list_lock);
461 
462 	/* bump our ref if we were not in the list before */
463 	if (!test_and_set_bit(RBIO_CACHE_BIT, &rbio->flags))
464 		refcount_inc(&rbio->refs);
465 
466 	if (!list_empty(&rbio->stripe_cache)){
467 		list_move(&rbio->stripe_cache, &table->stripe_cache);
468 	} else {
469 		list_add(&rbio->stripe_cache, &table->stripe_cache);
470 		table->cache_size += 1;
471 	}
472 
473 	spin_unlock(&rbio->bio_list_lock);
474 
475 	if (table->cache_size > RBIO_CACHE_SIZE) {
476 		struct btrfs_raid_bio *found;
477 
478 		found = list_entry(table->stripe_cache.prev,
479 				  struct btrfs_raid_bio,
480 				  stripe_cache);
481 
482 		if (found != rbio)
483 			__remove_rbio_from_cache(found);
484 	}
485 
486 	spin_unlock_irqrestore(&table->cache_lock, flags);
487 }
488 
489 /*
490  * helper function to run the xor_blocks api.  It is only
491  * able to do MAX_XOR_BLOCKS at a time, so we need to
492  * loop through.
493  */
494 static void run_xor(void **pages, int src_cnt, ssize_t len)
495 {
496 	int src_off = 0;
497 	int xor_src_cnt = 0;
498 	void *dest = pages[src_cnt];
499 
500 	while(src_cnt > 0) {
501 		xor_src_cnt = min(src_cnt, MAX_XOR_BLOCKS);
502 		xor_blocks(xor_src_cnt, len, dest, pages + src_off);
503 
504 		src_cnt -= xor_src_cnt;
505 		src_off += xor_src_cnt;
506 	}
507 }
508 
509 /*
510  * Returns true if the bio list inside this rbio covers an entire stripe (no
511  * rmw required).
512  */
513 static int rbio_is_full(struct btrfs_raid_bio *rbio)
514 {
515 	unsigned long flags;
516 	unsigned long size = rbio->bio_list_bytes;
517 	int ret = 1;
518 
519 	spin_lock_irqsave(&rbio->bio_list_lock, flags);
520 	if (size != rbio->nr_data * rbio->stripe_len)
521 		ret = 0;
522 	BUG_ON(size > rbio->nr_data * rbio->stripe_len);
523 	spin_unlock_irqrestore(&rbio->bio_list_lock, flags);
524 
525 	return ret;
526 }
527 
528 /*
529  * returns 1 if it is safe to merge two rbios together.
530  * The merging is safe if the two rbios correspond to
531  * the same stripe and if they are both going in the same
532  * direction (read vs write), and if neither one is
533  * locked for final IO
534  *
535  * The caller is responsible for locking such that
536  * rmw_locked is safe to test
537  */
538 static int rbio_can_merge(struct btrfs_raid_bio *last,
539 			  struct btrfs_raid_bio *cur)
540 {
541 	if (test_bit(RBIO_RMW_LOCKED_BIT, &last->flags) ||
542 	    test_bit(RBIO_RMW_LOCKED_BIT, &cur->flags))
543 		return 0;
544 
545 	/*
546 	 * we can't merge with cached rbios, since the
547 	 * idea is that when we merge the destination
548 	 * rbio is going to run our IO for us.  We can
549 	 * steal from cached rbios though, other functions
550 	 * handle that.
551 	 */
552 	if (test_bit(RBIO_CACHE_BIT, &last->flags) ||
553 	    test_bit(RBIO_CACHE_BIT, &cur->flags))
554 		return 0;
555 
556 	if (last->bbio->raid_map[0] !=
557 	    cur->bbio->raid_map[0])
558 		return 0;
559 
560 	/* we can't merge with different operations */
561 	if (last->operation != cur->operation)
562 		return 0;
563 	/*
564 	 * We've need read the full stripe from the drive.
565 	 * check and repair the parity and write the new results.
566 	 *
567 	 * We're not allowed to add any new bios to the
568 	 * bio list here, anyone else that wants to
569 	 * change this stripe needs to do their own rmw.
570 	 */
571 	if (last->operation == BTRFS_RBIO_PARITY_SCRUB)
572 		return 0;
573 
574 	if (last->operation == BTRFS_RBIO_REBUILD_MISSING)
575 		return 0;
576 
577 	if (last->operation == BTRFS_RBIO_READ_REBUILD) {
578 		int fa = last->faila;
579 		int fb = last->failb;
580 		int cur_fa = cur->faila;
581 		int cur_fb = cur->failb;
582 
583 		if (last->faila >= last->failb) {
584 			fa = last->failb;
585 			fb = last->faila;
586 		}
587 
588 		if (cur->faila >= cur->failb) {
589 			cur_fa = cur->failb;
590 			cur_fb = cur->faila;
591 		}
592 
593 		if (fa != cur_fa || fb != cur_fb)
594 			return 0;
595 	}
596 	return 1;
597 }
598 
599 static int rbio_stripe_page_index(struct btrfs_raid_bio *rbio, int stripe,
600 				  int index)
601 {
602 	return stripe * rbio->stripe_npages + index;
603 }
604 
605 /*
606  * these are just the pages from the rbio array, not from anything
607  * the FS sent down to us
608  */
609 static struct page *rbio_stripe_page(struct btrfs_raid_bio *rbio, int stripe,
610 				     int index)
611 {
612 	return rbio->stripe_pages[rbio_stripe_page_index(rbio, stripe, index)];
613 }
614 
615 /*
616  * helper to index into the pstripe
617  */
618 static struct page *rbio_pstripe_page(struct btrfs_raid_bio *rbio, int index)
619 {
620 	return rbio_stripe_page(rbio, rbio->nr_data, index);
621 }
622 
623 /*
624  * helper to index into the qstripe, returns null
625  * if there is no qstripe
626  */
627 static struct page *rbio_qstripe_page(struct btrfs_raid_bio *rbio, int index)
628 {
629 	if (rbio->nr_data + 1 == rbio->real_stripes)
630 		return NULL;
631 	return rbio_stripe_page(rbio, rbio->nr_data + 1, index);
632 }
633 
634 /*
635  * The first stripe in the table for a logical address
636  * has the lock.  rbios are added in one of three ways:
637  *
638  * 1) Nobody has the stripe locked yet.  The rbio is given
639  * the lock and 0 is returned.  The caller must start the IO
640  * themselves.
641  *
642  * 2) Someone has the stripe locked, but we're able to merge
643  * with the lock owner.  The rbio is freed and the IO will
644  * start automatically along with the existing rbio.  1 is returned.
645  *
646  * 3) Someone has the stripe locked, but we're not able to merge.
647  * The rbio is added to the lock owner's plug list, or merged into
648  * an rbio already on the plug list.  When the lock owner unlocks,
649  * the next rbio on the list is run and the IO is started automatically.
650  * 1 is returned
651  *
652  * If we return 0, the caller still owns the rbio and must continue with
653  * IO submission.  If we return 1, the caller must assume the rbio has
654  * already been freed.
655  */
656 static noinline int lock_stripe_add(struct btrfs_raid_bio *rbio)
657 {
658 	int bucket = rbio_bucket(rbio);
659 	struct btrfs_stripe_hash *h = rbio->fs_info->stripe_hash_table->table + bucket;
660 	struct btrfs_raid_bio *cur;
661 	struct btrfs_raid_bio *pending;
662 	unsigned long flags;
663 	struct btrfs_raid_bio *freeit = NULL;
664 	struct btrfs_raid_bio *cache_drop = NULL;
665 	int ret = 0;
666 
667 	spin_lock_irqsave(&h->lock, flags);
668 	list_for_each_entry(cur, &h->hash_list, hash_list) {
669 		if (cur->bbio->raid_map[0] == rbio->bbio->raid_map[0]) {
670 			spin_lock(&cur->bio_list_lock);
671 
672 			/* can we steal this cached rbio's pages? */
673 			if (bio_list_empty(&cur->bio_list) &&
674 			    list_empty(&cur->plug_list) &&
675 			    test_bit(RBIO_CACHE_BIT, &cur->flags) &&
676 			    !test_bit(RBIO_RMW_LOCKED_BIT, &cur->flags)) {
677 				list_del_init(&cur->hash_list);
678 				refcount_dec(&cur->refs);
679 
680 				steal_rbio(cur, rbio);
681 				cache_drop = cur;
682 				spin_unlock(&cur->bio_list_lock);
683 
684 				goto lockit;
685 			}
686 
687 			/* can we merge into the lock owner? */
688 			if (rbio_can_merge(cur, rbio)) {
689 				merge_rbio(cur, rbio);
690 				spin_unlock(&cur->bio_list_lock);
691 				freeit = rbio;
692 				ret = 1;
693 				goto out;
694 			}
695 
696 
697 			/*
698 			 * we couldn't merge with the running
699 			 * rbio, see if we can merge with the
700 			 * pending ones.  We don't have to
701 			 * check for rmw_locked because there
702 			 * is no way they are inside finish_rmw
703 			 * right now
704 			 */
705 			list_for_each_entry(pending, &cur->plug_list,
706 					    plug_list) {
707 				if (rbio_can_merge(pending, rbio)) {
708 					merge_rbio(pending, rbio);
709 					spin_unlock(&cur->bio_list_lock);
710 					freeit = rbio;
711 					ret = 1;
712 					goto out;
713 				}
714 			}
715 
716 			/* no merging, put us on the tail of the plug list,
717 			 * our rbio will be started with the currently
718 			 * running rbio unlocks
719 			 */
720 			list_add_tail(&rbio->plug_list, &cur->plug_list);
721 			spin_unlock(&cur->bio_list_lock);
722 			ret = 1;
723 			goto out;
724 		}
725 	}
726 lockit:
727 	refcount_inc(&rbio->refs);
728 	list_add(&rbio->hash_list, &h->hash_list);
729 out:
730 	spin_unlock_irqrestore(&h->lock, flags);
731 	if (cache_drop)
732 		remove_rbio_from_cache(cache_drop);
733 	if (freeit)
734 		__free_raid_bio(freeit);
735 	return ret;
736 }
737 
738 /*
739  * called as rmw or parity rebuild is completed.  If the plug list has more
740  * rbios waiting for this stripe, the next one on the list will be started
741  */
742 static noinline void unlock_stripe(struct btrfs_raid_bio *rbio)
743 {
744 	int bucket;
745 	struct btrfs_stripe_hash *h;
746 	unsigned long flags;
747 	int keep_cache = 0;
748 
749 	bucket = rbio_bucket(rbio);
750 	h = rbio->fs_info->stripe_hash_table->table + bucket;
751 
752 	if (list_empty(&rbio->plug_list))
753 		cache_rbio(rbio);
754 
755 	spin_lock_irqsave(&h->lock, flags);
756 	spin_lock(&rbio->bio_list_lock);
757 
758 	if (!list_empty(&rbio->hash_list)) {
759 		/*
760 		 * if we're still cached and there is no other IO
761 		 * to perform, just leave this rbio here for others
762 		 * to steal from later
763 		 */
764 		if (list_empty(&rbio->plug_list) &&
765 		    test_bit(RBIO_CACHE_BIT, &rbio->flags)) {
766 			keep_cache = 1;
767 			clear_bit(RBIO_RMW_LOCKED_BIT, &rbio->flags);
768 			BUG_ON(!bio_list_empty(&rbio->bio_list));
769 			goto done;
770 		}
771 
772 		list_del_init(&rbio->hash_list);
773 		refcount_dec(&rbio->refs);
774 
775 		/*
776 		 * we use the plug list to hold all the rbios
777 		 * waiting for the chance to lock this stripe.
778 		 * hand the lock over to one of them.
779 		 */
780 		if (!list_empty(&rbio->plug_list)) {
781 			struct btrfs_raid_bio *next;
782 			struct list_head *head = rbio->plug_list.next;
783 
784 			next = list_entry(head, struct btrfs_raid_bio,
785 					  plug_list);
786 
787 			list_del_init(&rbio->plug_list);
788 
789 			list_add(&next->hash_list, &h->hash_list);
790 			refcount_inc(&next->refs);
791 			spin_unlock(&rbio->bio_list_lock);
792 			spin_unlock_irqrestore(&h->lock, flags);
793 
794 			if (next->operation == BTRFS_RBIO_READ_REBUILD)
795 				start_async_work(next, read_rebuild_work);
796 			else if (next->operation == BTRFS_RBIO_REBUILD_MISSING) {
797 				steal_rbio(rbio, next);
798 				start_async_work(next, read_rebuild_work);
799 			} else if (next->operation == BTRFS_RBIO_WRITE) {
800 				steal_rbio(rbio, next);
801 				start_async_work(next, rmw_work);
802 			} else if (next->operation == BTRFS_RBIO_PARITY_SCRUB) {
803 				steal_rbio(rbio, next);
804 				start_async_work(next, scrub_parity_work);
805 			}
806 
807 			goto done_nolock;
808 		}
809 	}
810 done:
811 	spin_unlock(&rbio->bio_list_lock);
812 	spin_unlock_irqrestore(&h->lock, flags);
813 
814 done_nolock:
815 	if (!keep_cache)
816 		remove_rbio_from_cache(rbio);
817 }
818 
819 static void __free_raid_bio(struct btrfs_raid_bio *rbio)
820 {
821 	int i;
822 
823 	if (!refcount_dec_and_test(&rbio->refs))
824 		return;
825 
826 	WARN_ON(!list_empty(&rbio->stripe_cache));
827 	WARN_ON(!list_empty(&rbio->hash_list));
828 	WARN_ON(!bio_list_empty(&rbio->bio_list));
829 
830 	for (i = 0; i < rbio->nr_pages; i++) {
831 		if (rbio->stripe_pages[i]) {
832 			__free_page(rbio->stripe_pages[i]);
833 			rbio->stripe_pages[i] = NULL;
834 		}
835 	}
836 
837 	btrfs_put_bbio(rbio->bbio);
838 	kfree(rbio);
839 }
840 
841 static void rbio_endio_bio_list(struct bio *cur, blk_status_t err)
842 {
843 	struct bio *next;
844 
845 	while (cur) {
846 		next = cur->bi_next;
847 		cur->bi_next = NULL;
848 		cur->bi_status = err;
849 		bio_endio(cur);
850 		cur = next;
851 	}
852 }
853 
854 /*
855  * this frees the rbio and runs through all the bios in the
856  * bio_list and calls end_io on them
857  */
858 static void rbio_orig_end_io(struct btrfs_raid_bio *rbio, blk_status_t err)
859 {
860 	struct bio *cur = bio_list_get(&rbio->bio_list);
861 	struct bio *extra;
862 
863 	if (rbio->generic_bio_cnt)
864 		btrfs_bio_counter_sub(rbio->fs_info, rbio->generic_bio_cnt);
865 
866 	/*
867 	 * At this moment, rbio->bio_list is empty, however since rbio does not
868 	 * always have RBIO_RMW_LOCKED_BIT set and rbio is still linked on the
869 	 * hash list, rbio may be merged with others so that rbio->bio_list
870 	 * becomes non-empty.
871 	 * Once unlock_stripe() is done, rbio->bio_list will not be updated any
872 	 * more and we can call bio_endio() on all queued bios.
873 	 */
874 	unlock_stripe(rbio);
875 	extra = bio_list_get(&rbio->bio_list);
876 	__free_raid_bio(rbio);
877 
878 	rbio_endio_bio_list(cur, err);
879 	if (extra)
880 		rbio_endio_bio_list(extra, err);
881 }
882 
883 /*
884  * end io function used by finish_rmw.  When we finally
885  * get here, we've written a full stripe
886  */
887 static void raid_write_end_io(struct bio *bio)
888 {
889 	struct btrfs_raid_bio *rbio = bio->bi_private;
890 	blk_status_t err = bio->bi_status;
891 	int max_errors;
892 
893 	if (err)
894 		fail_bio_stripe(rbio, bio);
895 
896 	bio_put(bio);
897 
898 	if (!atomic_dec_and_test(&rbio->stripes_pending))
899 		return;
900 
901 	err = BLK_STS_OK;
902 
903 	/* OK, we have read all the stripes we need to. */
904 	max_errors = (rbio->operation == BTRFS_RBIO_PARITY_SCRUB) ?
905 		     0 : rbio->bbio->max_errors;
906 	if (atomic_read(&rbio->error) > max_errors)
907 		err = BLK_STS_IOERR;
908 
909 	rbio_orig_end_io(rbio, err);
910 }
911 
912 /*
913  * the read/modify/write code wants to use the original bio for
914  * any pages it included, and then use the rbio for everything
915  * else.  This function decides if a given index (stripe number)
916  * and page number in that stripe fall inside the original bio
917  * or the rbio.
918  *
919  * if you set bio_list_only, you'll get a NULL back for any ranges
920  * that are outside the bio_list
921  *
922  * This doesn't take any refs on anything, you get a bare page pointer
923  * and the caller must bump refs as required.
924  *
925  * You must call index_rbio_pages once before you can trust
926  * the answers from this function.
927  */
928 static struct page *page_in_rbio(struct btrfs_raid_bio *rbio,
929 				 int index, int pagenr, int bio_list_only)
930 {
931 	int chunk_page;
932 	struct page *p = NULL;
933 
934 	chunk_page = index * (rbio->stripe_len >> PAGE_SHIFT) + pagenr;
935 
936 	spin_lock_irq(&rbio->bio_list_lock);
937 	p = rbio->bio_pages[chunk_page];
938 	spin_unlock_irq(&rbio->bio_list_lock);
939 
940 	if (p || bio_list_only)
941 		return p;
942 
943 	return rbio->stripe_pages[chunk_page];
944 }
945 
946 /*
947  * number of pages we need for the entire stripe across all the
948  * drives
949  */
950 static unsigned long rbio_nr_pages(unsigned long stripe_len, int nr_stripes)
951 {
952 	return DIV_ROUND_UP(stripe_len, PAGE_SIZE) * nr_stripes;
953 }
954 
955 /*
956  * allocation and initial setup for the btrfs_raid_bio.  Not
957  * this does not allocate any pages for rbio->pages.
958  */
959 static struct btrfs_raid_bio *alloc_rbio(struct btrfs_fs_info *fs_info,
960 					 struct btrfs_bio *bbio,
961 					 u64 stripe_len)
962 {
963 	struct btrfs_raid_bio *rbio;
964 	int nr_data = 0;
965 	int real_stripes = bbio->num_stripes - bbio->num_tgtdevs;
966 	int num_pages = rbio_nr_pages(stripe_len, real_stripes);
967 	int stripe_npages = DIV_ROUND_UP(stripe_len, PAGE_SIZE);
968 	void *p;
969 
970 	rbio = kzalloc(sizeof(*rbio) +
971 		       sizeof(*rbio->stripe_pages) * num_pages +
972 		       sizeof(*rbio->bio_pages) * num_pages +
973 		       sizeof(*rbio->finish_pointers) * real_stripes +
974 		       sizeof(*rbio->dbitmap) * BITS_TO_LONGS(stripe_npages) +
975 		       sizeof(*rbio->finish_pbitmap) *
976 				BITS_TO_LONGS(stripe_npages),
977 		       GFP_NOFS);
978 	if (!rbio)
979 		return ERR_PTR(-ENOMEM);
980 
981 	bio_list_init(&rbio->bio_list);
982 	INIT_LIST_HEAD(&rbio->plug_list);
983 	spin_lock_init(&rbio->bio_list_lock);
984 	INIT_LIST_HEAD(&rbio->stripe_cache);
985 	INIT_LIST_HEAD(&rbio->hash_list);
986 	rbio->bbio = bbio;
987 	rbio->fs_info = fs_info;
988 	rbio->stripe_len = stripe_len;
989 	rbio->nr_pages = num_pages;
990 	rbio->real_stripes = real_stripes;
991 	rbio->stripe_npages = stripe_npages;
992 	rbio->faila = -1;
993 	rbio->failb = -1;
994 	refcount_set(&rbio->refs, 1);
995 	atomic_set(&rbio->error, 0);
996 	atomic_set(&rbio->stripes_pending, 0);
997 
998 	/*
999 	 * the stripe_pages, bio_pages, etc arrays point to the extra
1000 	 * memory we allocated past the end of the rbio
1001 	 */
1002 	p = rbio + 1;
1003 #define CONSUME_ALLOC(ptr, count)	do {				\
1004 		ptr = p;						\
1005 		p = (unsigned char *)p + sizeof(*(ptr)) * (count);	\
1006 	} while (0)
1007 	CONSUME_ALLOC(rbio->stripe_pages, num_pages);
1008 	CONSUME_ALLOC(rbio->bio_pages, num_pages);
1009 	CONSUME_ALLOC(rbio->finish_pointers, real_stripes);
1010 	CONSUME_ALLOC(rbio->dbitmap, BITS_TO_LONGS(stripe_npages));
1011 	CONSUME_ALLOC(rbio->finish_pbitmap, BITS_TO_LONGS(stripe_npages));
1012 #undef  CONSUME_ALLOC
1013 
1014 	if (bbio->map_type & BTRFS_BLOCK_GROUP_RAID5)
1015 		nr_data = real_stripes - 1;
1016 	else if (bbio->map_type & BTRFS_BLOCK_GROUP_RAID6)
1017 		nr_data = real_stripes - 2;
1018 	else
1019 		BUG();
1020 
1021 	rbio->nr_data = nr_data;
1022 	return rbio;
1023 }
1024 
1025 /* allocate pages for all the stripes in the bio, including parity */
1026 static int alloc_rbio_pages(struct btrfs_raid_bio *rbio)
1027 {
1028 	int i;
1029 	struct page *page;
1030 
1031 	for (i = 0; i < rbio->nr_pages; i++) {
1032 		if (rbio->stripe_pages[i])
1033 			continue;
1034 		page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
1035 		if (!page)
1036 			return -ENOMEM;
1037 		rbio->stripe_pages[i] = page;
1038 	}
1039 	return 0;
1040 }
1041 
1042 /* only allocate pages for p/q stripes */
1043 static int alloc_rbio_parity_pages(struct btrfs_raid_bio *rbio)
1044 {
1045 	int i;
1046 	struct page *page;
1047 
1048 	i = rbio_stripe_page_index(rbio, rbio->nr_data, 0);
1049 
1050 	for (; i < rbio->nr_pages; i++) {
1051 		if (rbio->stripe_pages[i])
1052 			continue;
1053 		page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
1054 		if (!page)
1055 			return -ENOMEM;
1056 		rbio->stripe_pages[i] = page;
1057 	}
1058 	return 0;
1059 }
1060 
1061 /*
1062  * add a single page from a specific stripe into our list of bios for IO
1063  * this will try to merge into existing bios if possible, and returns
1064  * zero if all went well.
1065  */
1066 static int rbio_add_io_page(struct btrfs_raid_bio *rbio,
1067 			    struct bio_list *bio_list,
1068 			    struct page *page,
1069 			    int stripe_nr,
1070 			    unsigned long page_index,
1071 			    unsigned long bio_max_len)
1072 {
1073 	struct bio *last = bio_list->tail;
1074 	u64 last_end = 0;
1075 	int ret;
1076 	struct bio *bio;
1077 	struct btrfs_bio_stripe *stripe;
1078 	u64 disk_start;
1079 
1080 	stripe = &rbio->bbio->stripes[stripe_nr];
1081 	disk_start = stripe->physical + (page_index << PAGE_SHIFT);
1082 
1083 	/* if the device is missing, just fail this stripe */
1084 	if (!stripe->dev->bdev)
1085 		return fail_rbio_index(rbio, stripe_nr);
1086 
1087 	/* see if we can add this page onto our existing bio */
1088 	if (last) {
1089 		last_end = (u64)last->bi_iter.bi_sector << 9;
1090 		last_end += last->bi_iter.bi_size;
1091 
1092 		/*
1093 		 * we can't merge these if they are from different
1094 		 * devices or if they are not contiguous
1095 		 */
1096 		if (last_end == disk_start && stripe->dev->bdev &&
1097 		    !last->bi_status &&
1098 		    last->bi_disk == stripe->dev->bdev->bd_disk &&
1099 		    last->bi_partno == stripe->dev->bdev->bd_partno) {
1100 			ret = bio_add_page(last, page, PAGE_SIZE, 0);
1101 			if (ret == PAGE_SIZE)
1102 				return 0;
1103 		}
1104 	}
1105 
1106 	/* put a new bio on the list */
1107 	bio = btrfs_io_bio_alloc(bio_max_len >> PAGE_SHIFT ?: 1);
1108 	bio->bi_iter.bi_size = 0;
1109 	bio_set_dev(bio, stripe->dev->bdev);
1110 	bio->bi_iter.bi_sector = disk_start >> 9;
1111 
1112 	bio_add_page(bio, page, PAGE_SIZE, 0);
1113 	bio_list_add(bio_list, bio);
1114 	return 0;
1115 }
1116 
1117 /*
1118  * while we're doing the read/modify/write cycle, we could
1119  * have errors in reading pages off the disk.  This checks
1120  * for errors and if we're not able to read the page it'll
1121  * trigger parity reconstruction.  The rmw will be finished
1122  * after we've reconstructed the failed stripes
1123  */
1124 static void validate_rbio_for_rmw(struct btrfs_raid_bio *rbio)
1125 {
1126 	if (rbio->faila >= 0 || rbio->failb >= 0) {
1127 		BUG_ON(rbio->faila == rbio->real_stripes - 1);
1128 		__raid56_parity_recover(rbio);
1129 	} else {
1130 		finish_rmw(rbio);
1131 	}
1132 }
1133 
1134 /*
1135  * helper function to walk our bio list and populate the bio_pages array with
1136  * the result.  This seems expensive, but it is faster than constantly
1137  * searching through the bio list as we setup the IO in finish_rmw or stripe
1138  * reconstruction.
1139  *
1140  * This must be called before you trust the answers from page_in_rbio
1141  */
1142 static void index_rbio_pages(struct btrfs_raid_bio *rbio)
1143 {
1144 	struct bio *bio;
1145 	u64 start;
1146 	unsigned long stripe_offset;
1147 	unsigned long page_index;
1148 
1149 	spin_lock_irq(&rbio->bio_list_lock);
1150 	bio_list_for_each(bio, &rbio->bio_list) {
1151 		struct bio_vec bvec;
1152 		struct bvec_iter iter;
1153 		int i = 0;
1154 
1155 		start = (u64)bio->bi_iter.bi_sector << 9;
1156 		stripe_offset = start - rbio->bbio->raid_map[0];
1157 		page_index = stripe_offset >> PAGE_SHIFT;
1158 
1159 		if (bio_flagged(bio, BIO_CLONED))
1160 			bio->bi_iter = btrfs_io_bio(bio)->iter;
1161 
1162 		bio_for_each_segment(bvec, bio, iter) {
1163 			rbio->bio_pages[page_index + i] = bvec.bv_page;
1164 			i++;
1165 		}
1166 	}
1167 	spin_unlock_irq(&rbio->bio_list_lock);
1168 }
1169 
1170 /*
1171  * this is called from one of two situations.  We either
1172  * have a full stripe from the higher layers, or we've read all
1173  * the missing bits off disk.
1174  *
1175  * This will calculate the parity and then send down any
1176  * changed blocks.
1177  */
1178 static noinline void finish_rmw(struct btrfs_raid_bio *rbio)
1179 {
1180 	struct btrfs_bio *bbio = rbio->bbio;
1181 	void **pointers = rbio->finish_pointers;
1182 	int nr_data = rbio->nr_data;
1183 	int stripe;
1184 	int pagenr;
1185 	int p_stripe = -1;
1186 	int q_stripe = -1;
1187 	struct bio_list bio_list;
1188 	struct bio *bio;
1189 	int ret;
1190 
1191 	bio_list_init(&bio_list);
1192 
1193 	if (rbio->real_stripes - rbio->nr_data == 1) {
1194 		p_stripe = rbio->real_stripes - 1;
1195 	} else if (rbio->real_stripes - rbio->nr_data == 2) {
1196 		p_stripe = rbio->real_stripes - 2;
1197 		q_stripe = rbio->real_stripes - 1;
1198 	} else {
1199 		BUG();
1200 	}
1201 
1202 	/* at this point we either have a full stripe,
1203 	 * or we've read the full stripe from the drive.
1204 	 * recalculate the parity and write the new results.
1205 	 *
1206 	 * We're not allowed to add any new bios to the
1207 	 * bio list here, anyone else that wants to
1208 	 * change this stripe needs to do their own rmw.
1209 	 */
1210 	spin_lock_irq(&rbio->bio_list_lock);
1211 	set_bit(RBIO_RMW_LOCKED_BIT, &rbio->flags);
1212 	spin_unlock_irq(&rbio->bio_list_lock);
1213 
1214 	atomic_set(&rbio->error, 0);
1215 
1216 	/*
1217 	 * now that we've set rmw_locked, run through the
1218 	 * bio list one last time and map the page pointers
1219 	 *
1220 	 * We don't cache full rbios because we're assuming
1221 	 * the higher layers are unlikely to use this area of
1222 	 * the disk again soon.  If they do use it again,
1223 	 * hopefully they will send another full bio.
1224 	 */
1225 	index_rbio_pages(rbio);
1226 	if (!rbio_is_full(rbio))
1227 		cache_rbio_pages(rbio);
1228 	else
1229 		clear_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
1230 
1231 	for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
1232 		struct page *p;
1233 		/* first collect one page from each data stripe */
1234 		for (stripe = 0; stripe < nr_data; stripe++) {
1235 			p = page_in_rbio(rbio, stripe, pagenr, 0);
1236 			pointers[stripe] = kmap(p);
1237 		}
1238 
1239 		/* then add the parity stripe */
1240 		p = rbio_pstripe_page(rbio, pagenr);
1241 		SetPageUptodate(p);
1242 		pointers[stripe++] = kmap(p);
1243 
1244 		if (q_stripe != -1) {
1245 
1246 			/*
1247 			 * raid6, add the qstripe and call the
1248 			 * library function to fill in our p/q
1249 			 */
1250 			p = rbio_qstripe_page(rbio, pagenr);
1251 			SetPageUptodate(p);
1252 			pointers[stripe++] = kmap(p);
1253 
1254 			raid6_call.gen_syndrome(rbio->real_stripes, PAGE_SIZE,
1255 						pointers);
1256 		} else {
1257 			/* raid5 */
1258 			copy_page(pointers[nr_data], pointers[0]);
1259 			run_xor(pointers + 1, nr_data - 1, PAGE_SIZE);
1260 		}
1261 
1262 
1263 		for (stripe = 0; stripe < rbio->real_stripes; stripe++)
1264 			kunmap(page_in_rbio(rbio, stripe, pagenr, 0));
1265 	}
1266 
1267 	/*
1268 	 * time to start writing.  Make bios for everything from the
1269 	 * higher layers (the bio_list in our rbio) and our p/q.  Ignore
1270 	 * everything else.
1271 	 */
1272 	for (stripe = 0; stripe < rbio->real_stripes; stripe++) {
1273 		for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
1274 			struct page *page;
1275 			if (stripe < rbio->nr_data) {
1276 				page = page_in_rbio(rbio, stripe, pagenr, 1);
1277 				if (!page)
1278 					continue;
1279 			} else {
1280 			       page = rbio_stripe_page(rbio, stripe, pagenr);
1281 			}
1282 
1283 			ret = rbio_add_io_page(rbio, &bio_list,
1284 				       page, stripe, pagenr, rbio->stripe_len);
1285 			if (ret)
1286 				goto cleanup;
1287 		}
1288 	}
1289 
1290 	if (likely(!bbio->num_tgtdevs))
1291 		goto write_data;
1292 
1293 	for (stripe = 0; stripe < rbio->real_stripes; stripe++) {
1294 		if (!bbio->tgtdev_map[stripe])
1295 			continue;
1296 
1297 		for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
1298 			struct page *page;
1299 			if (stripe < rbio->nr_data) {
1300 				page = page_in_rbio(rbio, stripe, pagenr, 1);
1301 				if (!page)
1302 					continue;
1303 			} else {
1304 			       page = rbio_stripe_page(rbio, stripe, pagenr);
1305 			}
1306 
1307 			ret = rbio_add_io_page(rbio, &bio_list, page,
1308 					       rbio->bbio->tgtdev_map[stripe],
1309 					       pagenr, rbio->stripe_len);
1310 			if (ret)
1311 				goto cleanup;
1312 		}
1313 	}
1314 
1315 write_data:
1316 	atomic_set(&rbio->stripes_pending, bio_list_size(&bio_list));
1317 	BUG_ON(atomic_read(&rbio->stripes_pending) == 0);
1318 
1319 	while (1) {
1320 		bio = bio_list_pop(&bio_list);
1321 		if (!bio)
1322 			break;
1323 
1324 		bio->bi_private = rbio;
1325 		bio->bi_end_io = raid_write_end_io;
1326 		bio->bi_opf = REQ_OP_WRITE;
1327 
1328 		submit_bio(bio);
1329 	}
1330 	return;
1331 
1332 cleanup:
1333 	rbio_orig_end_io(rbio, BLK_STS_IOERR);
1334 
1335 	while ((bio = bio_list_pop(&bio_list)))
1336 		bio_put(bio);
1337 }
1338 
1339 /*
1340  * helper to find the stripe number for a given bio.  Used to figure out which
1341  * stripe has failed.  This expects the bio to correspond to a physical disk,
1342  * so it looks up based on physical sector numbers.
1343  */
1344 static int find_bio_stripe(struct btrfs_raid_bio *rbio,
1345 			   struct bio *bio)
1346 {
1347 	u64 physical = bio->bi_iter.bi_sector;
1348 	u64 stripe_start;
1349 	int i;
1350 	struct btrfs_bio_stripe *stripe;
1351 
1352 	physical <<= 9;
1353 
1354 	for (i = 0; i < rbio->bbio->num_stripes; i++) {
1355 		stripe = &rbio->bbio->stripes[i];
1356 		stripe_start = stripe->physical;
1357 		if (physical >= stripe_start &&
1358 		    physical < stripe_start + rbio->stripe_len &&
1359 		    stripe->dev->bdev &&
1360 		    bio->bi_disk == stripe->dev->bdev->bd_disk &&
1361 		    bio->bi_partno == stripe->dev->bdev->bd_partno) {
1362 			return i;
1363 		}
1364 	}
1365 	return -1;
1366 }
1367 
1368 /*
1369  * helper to find the stripe number for a given
1370  * bio (before mapping).  Used to figure out which stripe has
1371  * failed.  This looks up based on logical block numbers.
1372  */
1373 static int find_logical_bio_stripe(struct btrfs_raid_bio *rbio,
1374 				   struct bio *bio)
1375 {
1376 	u64 logical = bio->bi_iter.bi_sector;
1377 	u64 stripe_start;
1378 	int i;
1379 
1380 	logical <<= 9;
1381 
1382 	for (i = 0; i < rbio->nr_data; i++) {
1383 		stripe_start = rbio->bbio->raid_map[i];
1384 		if (logical >= stripe_start &&
1385 		    logical < stripe_start + rbio->stripe_len) {
1386 			return i;
1387 		}
1388 	}
1389 	return -1;
1390 }
1391 
1392 /*
1393  * returns -EIO if we had too many failures
1394  */
1395 static int fail_rbio_index(struct btrfs_raid_bio *rbio, int failed)
1396 {
1397 	unsigned long flags;
1398 	int ret = 0;
1399 
1400 	spin_lock_irqsave(&rbio->bio_list_lock, flags);
1401 
1402 	/* we already know this stripe is bad, move on */
1403 	if (rbio->faila == failed || rbio->failb == failed)
1404 		goto out;
1405 
1406 	if (rbio->faila == -1) {
1407 		/* first failure on this rbio */
1408 		rbio->faila = failed;
1409 		atomic_inc(&rbio->error);
1410 	} else if (rbio->failb == -1) {
1411 		/* second failure on this rbio */
1412 		rbio->failb = failed;
1413 		atomic_inc(&rbio->error);
1414 	} else {
1415 		ret = -EIO;
1416 	}
1417 out:
1418 	spin_unlock_irqrestore(&rbio->bio_list_lock, flags);
1419 
1420 	return ret;
1421 }
1422 
1423 /*
1424  * helper to fail a stripe based on a physical disk
1425  * bio.
1426  */
1427 static int fail_bio_stripe(struct btrfs_raid_bio *rbio,
1428 			   struct bio *bio)
1429 {
1430 	int failed = find_bio_stripe(rbio, bio);
1431 
1432 	if (failed < 0)
1433 		return -EIO;
1434 
1435 	return fail_rbio_index(rbio, failed);
1436 }
1437 
1438 /*
1439  * this sets each page in the bio uptodate.  It should only be used on private
1440  * rbio pages, nothing that comes in from the higher layers
1441  */
1442 static void set_bio_pages_uptodate(struct bio *bio)
1443 {
1444 	struct bio_vec *bvec;
1445 	int i;
1446 
1447 	ASSERT(!bio_flagged(bio, BIO_CLONED));
1448 
1449 	bio_for_each_segment_all(bvec, bio, i)
1450 		SetPageUptodate(bvec->bv_page);
1451 }
1452 
1453 /*
1454  * end io for the read phase of the rmw cycle.  All the bios here are physical
1455  * stripe bios we've read from the disk so we can recalculate the parity of the
1456  * stripe.
1457  *
1458  * This will usually kick off finish_rmw once all the bios are read in, but it
1459  * may trigger parity reconstruction if we had any errors along the way
1460  */
1461 static void raid_rmw_end_io(struct bio *bio)
1462 {
1463 	struct btrfs_raid_bio *rbio = bio->bi_private;
1464 
1465 	if (bio->bi_status)
1466 		fail_bio_stripe(rbio, bio);
1467 	else
1468 		set_bio_pages_uptodate(bio);
1469 
1470 	bio_put(bio);
1471 
1472 	if (!atomic_dec_and_test(&rbio->stripes_pending))
1473 		return;
1474 
1475 	if (atomic_read(&rbio->error) > rbio->bbio->max_errors)
1476 		goto cleanup;
1477 
1478 	/*
1479 	 * this will normally call finish_rmw to start our write
1480 	 * but if there are any failed stripes we'll reconstruct
1481 	 * from parity first
1482 	 */
1483 	validate_rbio_for_rmw(rbio);
1484 	return;
1485 
1486 cleanup:
1487 
1488 	rbio_orig_end_io(rbio, BLK_STS_IOERR);
1489 }
1490 
1491 /*
1492  * the stripe must be locked by the caller.  It will
1493  * unlock after all the writes are done
1494  */
1495 static int raid56_rmw_stripe(struct btrfs_raid_bio *rbio)
1496 {
1497 	int bios_to_read = 0;
1498 	struct bio_list bio_list;
1499 	int ret;
1500 	int pagenr;
1501 	int stripe;
1502 	struct bio *bio;
1503 
1504 	bio_list_init(&bio_list);
1505 
1506 	ret = alloc_rbio_pages(rbio);
1507 	if (ret)
1508 		goto cleanup;
1509 
1510 	index_rbio_pages(rbio);
1511 
1512 	atomic_set(&rbio->error, 0);
1513 	/*
1514 	 * build a list of bios to read all the missing parts of this
1515 	 * stripe
1516 	 */
1517 	for (stripe = 0; stripe < rbio->nr_data; stripe++) {
1518 		for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
1519 			struct page *page;
1520 			/*
1521 			 * we want to find all the pages missing from
1522 			 * the rbio and read them from the disk.  If
1523 			 * page_in_rbio finds a page in the bio list
1524 			 * we don't need to read it off the stripe.
1525 			 */
1526 			page = page_in_rbio(rbio, stripe, pagenr, 1);
1527 			if (page)
1528 				continue;
1529 
1530 			page = rbio_stripe_page(rbio, stripe, pagenr);
1531 			/*
1532 			 * the bio cache may have handed us an uptodate
1533 			 * page.  If so, be happy and use it
1534 			 */
1535 			if (PageUptodate(page))
1536 				continue;
1537 
1538 			ret = rbio_add_io_page(rbio, &bio_list, page,
1539 				       stripe, pagenr, rbio->stripe_len);
1540 			if (ret)
1541 				goto cleanup;
1542 		}
1543 	}
1544 
1545 	bios_to_read = bio_list_size(&bio_list);
1546 	if (!bios_to_read) {
1547 		/*
1548 		 * this can happen if others have merged with
1549 		 * us, it means there is nothing left to read.
1550 		 * But if there are missing devices it may not be
1551 		 * safe to do the full stripe write yet.
1552 		 */
1553 		goto finish;
1554 	}
1555 
1556 	/*
1557 	 * the bbio may be freed once we submit the last bio.  Make sure
1558 	 * not to touch it after that
1559 	 */
1560 	atomic_set(&rbio->stripes_pending, bios_to_read);
1561 	while (1) {
1562 		bio = bio_list_pop(&bio_list);
1563 		if (!bio)
1564 			break;
1565 
1566 		bio->bi_private = rbio;
1567 		bio->bi_end_io = raid_rmw_end_io;
1568 		bio->bi_opf = REQ_OP_READ;
1569 
1570 		btrfs_bio_wq_end_io(rbio->fs_info, bio, BTRFS_WQ_ENDIO_RAID56);
1571 
1572 		submit_bio(bio);
1573 	}
1574 	/* the actual write will happen once the reads are done */
1575 	return 0;
1576 
1577 cleanup:
1578 	rbio_orig_end_io(rbio, BLK_STS_IOERR);
1579 
1580 	while ((bio = bio_list_pop(&bio_list)))
1581 		bio_put(bio);
1582 
1583 	return -EIO;
1584 
1585 finish:
1586 	validate_rbio_for_rmw(rbio);
1587 	return 0;
1588 }
1589 
1590 /*
1591  * if the upper layers pass in a full stripe, we thank them by only allocating
1592  * enough pages to hold the parity, and sending it all down quickly.
1593  */
1594 static int full_stripe_write(struct btrfs_raid_bio *rbio)
1595 {
1596 	int ret;
1597 
1598 	ret = alloc_rbio_parity_pages(rbio);
1599 	if (ret) {
1600 		__free_raid_bio(rbio);
1601 		return ret;
1602 	}
1603 
1604 	ret = lock_stripe_add(rbio);
1605 	if (ret == 0)
1606 		finish_rmw(rbio);
1607 	return 0;
1608 }
1609 
1610 /*
1611  * partial stripe writes get handed over to async helpers.
1612  * We're really hoping to merge a few more writes into this
1613  * rbio before calculating new parity
1614  */
1615 static int partial_stripe_write(struct btrfs_raid_bio *rbio)
1616 {
1617 	int ret;
1618 
1619 	ret = lock_stripe_add(rbio);
1620 	if (ret == 0)
1621 		start_async_work(rbio, rmw_work);
1622 	return 0;
1623 }
1624 
1625 /*
1626  * sometimes while we were reading from the drive to
1627  * recalculate parity, enough new bios come into create
1628  * a full stripe.  So we do a check here to see if we can
1629  * go directly to finish_rmw
1630  */
1631 static int __raid56_parity_write(struct btrfs_raid_bio *rbio)
1632 {
1633 	/* head off into rmw land if we don't have a full stripe */
1634 	if (!rbio_is_full(rbio))
1635 		return partial_stripe_write(rbio);
1636 	return full_stripe_write(rbio);
1637 }
1638 
1639 /*
1640  * We use plugging call backs to collect full stripes.
1641  * Any time we get a partial stripe write while plugged
1642  * we collect it into a list.  When the unplug comes down,
1643  * we sort the list by logical block number and merge
1644  * everything we can into the same rbios
1645  */
1646 struct btrfs_plug_cb {
1647 	struct blk_plug_cb cb;
1648 	struct btrfs_fs_info *info;
1649 	struct list_head rbio_list;
1650 	struct btrfs_work work;
1651 };
1652 
1653 /*
1654  * rbios on the plug list are sorted for easier merging.
1655  */
1656 static int plug_cmp(void *priv, struct list_head *a, struct list_head *b)
1657 {
1658 	struct btrfs_raid_bio *ra = container_of(a, struct btrfs_raid_bio,
1659 						 plug_list);
1660 	struct btrfs_raid_bio *rb = container_of(b, struct btrfs_raid_bio,
1661 						 plug_list);
1662 	u64 a_sector = ra->bio_list.head->bi_iter.bi_sector;
1663 	u64 b_sector = rb->bio_list.head->bi_iter.bi_sector;
1664 
1665 	if (a_sector < b_sector)
1666 		return -1;
1667 	if (a_sector > b_sector)
1668 		return 1;
1669 	return 0;
1670 }
1671 
1672 static void run_plug(struct btrfs_plug_cb *plug)
1673 {
1674 	struct btrfs_raid_bio *cur;
1675 	struct btrfs_raid_bio *last = NULL;
1676 
1677 	/*
1678 	 * sort our plug list then try to merge
1679 	 * everything we can in hopes of creating full
1680 	 * stripes.
1681 	 */
1682 	list_sort(NULL, &plug->rbio_list, plug_cmp);
1683 	while (!list_empty(&plug->rbio_list)) {
1684 		cur = list_entry(plug->rbio_list.next,
1685 				 struct btrfs_raid_bio, plug_list);
1686 		list_del_init(&cur->plug_list);
1687 
1688 		if (rbio_is_full(cur)) {
1689 			int ret;
1690 
1691 			/* we have a full stripe, send it down */
1692 			ret = full_stripe_write(cur);
1693 			BUG_ON(ret);
1694 			continue;
1695 		}
1696 		if (last) {
1697 			if (rbio_can_merge(last, cur)) {
1698 				merge_rbio(last, cur);
1699 				__free_raid_bio(cur);
1700 				continue;
1701 
1702 			}
1703 			__raid56_parity_write(last);
1704 		}
1705 		last = cur;
1706 	}
1707 	if (last) {
1708 		__raid56_parity_write(last);
1709 	}
1710 	kfree(plug);
1711 }
1712 
1713 /*
1714  * if the unplug comes from schedule, we have to push the
1715  * work off to a helper thread
1716  */
1717 static void unplug_work(struct btrfs_work *work)
1718 {
1719 	struct btrfs_plug_cb *plug;
1720 	plug = container_of(work, struct btrfs_plug_cb, work);
1721 	run_plug(plug);
1722 }
1723 
1724 static void btrfs_raid_unplug(struct blk_plug_cb *cb, bool from_schedule)
1725 {
1726 	struct btrfs_plug_cb *plug;
1727 	plug = container_of(cb, struct btrfs_plug_cb, cb);
1728 
1729 	if (from_schedule) {
1730 		btrfs_init_work(&plug->work, btrfs_rmw_helper,
1731 				unplug_work, NULL, NULL);
1732 		btrfs_queue_work(plug->info->rmw_workers,
1733 				 &plug->work);
1734 		return;
1735 	}
1736 	run_plug(plug);
1737 }
1738 
1739 /*
1740  * our main entry point for writes from the rest of the FS.
1741  */
1742 int raid56_parity_write(struct btrfs_fs_info *fs_info, struct bio *bio,
1743 			struct btrfs_bio *bbio, u64 stripe_len)
1744 {
1745 	struct btrfs_raid_bio *rbio;
1746 	struct btrfs_plug_cb *plug = NULL;
1747 	struct blk_plug_cb *cb;
1748 	int ret;
1749 
1750 	rbio = alloc_rbio(fs_info, bbio, stripe_len);
1751 	if (IS_ERR(rbio)) {
1752 		btrfs_put_bbio(bbio);
1753 		return PTR_ERR(rbio);
1754 	}
1755 	bio_list_add(&rbio->bio_list, bio);
1756 	rbio->bio_list_bytes = bio->bi_iter.bi_size;
1757 	rbio->operation = BTRFS_RBIO_WRITE;
1758 
1759 	btrfs_bio_counter_inc_noblocked(fs_info);
1760 	rbio->generic_bio_cnt = 1;
1761 
1762 	/*
1763 	 * don't plug on full rbios, just get them out the door
1764 	 * as quickly as we can
1765 	 */
1766 	if (rbio_is_full(rbio)) {
1767 		ret = full_stripe_write(rbio);
1768 		if (ret)
1769 			btrfs_bio_counter_dec(fs_info);
1770 		return ret;
1771 	}
1772 
1773 	cb = blk_check_plugged(btrfs_raid_unplug, fs_info, sizeof(*plug));
1774 	if (cb) {
1775 		plug = container_of(cb, struct btrfs_plug_cb, cb);
1776 		if (!plug->info) {
1777 			plug->info = fs_info;
1778 			INIT_LIST_HEAD(&plug->rbio_list);
1779 		}
1780 		list_add_tail(&rbio->plug_list, &plug->rbio_list);
1781 		ret = 0;
1782 	} else {
1783 		ret = __raid56_parity_write(rbio);
1784 		if (ret)
1785 			btrfs_bio_counter_dec(fs_info);
1786 	}
1787 	return ret;
1788 }
1789 
1790 /*
1791  * all parity reconstruction happens here.  We've read in everything
1792  * we can find from the drives and this does the heavy lifting of
1793  * sorting the good from the bad.
1794  */
1795 static void __raid_recover_end_io(struct btrfs_raid_bio *rbio)
1796 {
1797 	int pagenr, stripe;
1798 	void **pointers;
1799 	int faila = -1, failb = -1;
1800 	struct page *page;
1801 	blk_status_t err;
1802 	int i;
1803 
1804 	pointers = kcalloc(rbio->real_stripes, sizeof(void *), GFP_NOFS);
1805 	if (!pointers) {
1806 		err = BLK_STS_RESOURCE;
1807 		goto cleanup_io;
1808 	}
1809 
1810 	faila = rbio->faila;
1811 	failb = rbio->failb;
1812 
1813 	if (rbio->operation == BTRFS_RBIO_READ_REBUILD ||
1814 	    rbio->operation == BTRFS_RBIO_REBUILD_MISSING) {
1815 		spin_lock_irq(&rbio->bio_list_lock);
1816 		set_bit(RBIO_RMW_LOCKED_BIT, &rbio->flags);
1817 		spin_unlock_irq(&rbio->bio_list_lock);
1818 	}
1819 
1820 	index_rbio_pages(rbio);
1821 
1822 	for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
1823 		/*
1824 		 * Now we just use bitmap to mark the horizontal stripes in
1825 		 * which we have data when doing parity scrub.
1826 		 */
1827 		if (rbio->operation == BTRFS_RBIO_PARITY_SCRUB &&
1828 		    !test_bit(pagenr, rbio->dbitmap))
1829 			continue;
1830 
1831 		/* setup our array of pointers with pages
1832 		 * from each stripe
1833 		 */
1834 		for (stripe = 0; stripe < rbio->real_stripes; stripe++) {
1835 			/*
1836 			 * if we're rebuilding a read, we have to use
1837 			 * pages from the bio list
1838 			 */
1839 			if ((rbio->operation == BTRFS_RBIO_READ_REBUILD ||
1840 			     rbio->operation == BTRFS_RBIO_REBUILD_MISSING) &&
1841 			    (stripe == faila || stripe == failb)) {
1842 				page = page_in_rbio(rbio, stripe, pagenr, 0);
1843 			} else {
1844 				page = rbio_stripe_page(rbio, stripe, pagenr);
1845 			}
1846 			pointers[stripe] = kmap(page);
1847 		}
1848 
1849 		/* all raid6 handling here */
1850 		if (rbio->bbio->map_type & BTRFS_BLOCK_GROUP_RAID6) {
1851 			/*
1852 			 * single failure, rebuild from parity raid5
1853 			 * style
1854 			 */
1855 			if (failb < 0) {
1856 				if (faila == rbio->nr_data) {
1857 					/*
1858 					 * Just the P stripe has failed, without
1859 					 * a bad data or Q stripe.
1860 					 * TODO, we should redo the xor here.
1861 					 */
1862 					err = BLK_STS_IOERR;
1863 					goto cleanup;
1864 				}
1865 				/*
1866 				 * a single failure in raid6 is rebuilt
1867 				 * in the pstripe code below
1868 				 */
1869 				goto pstripe;
1870 			}
1871 
1872 			/* make sure our ps and qs are in order */
1873 			if (faila > failb) {
1874 				int tmp = failb;
1875 				failb = faila;
1876 				faila = tmp;
1877 			}
1878 
1879 			/* if the q stripe is failed, do a pstripe reconstruction
1880 			 * from the xors.
1881 			 * If both the q stripe and the P stripe are failed, we're
1882 			 * here due to a crc mismatch and we can't give them the
1883 			 * data they want
1884 			 */
1885 			if (rbio->bbio->raid_map[failb] == RAID6_Q_STRIPE) {
1886 				if (rbio->bbio->raid_map[faila] ==
1887 				    RAID5_P_STRIPE) {
1888 					err = BLK_STS_IOERR;
1889 					goto cleanup;
1890 				}
1891 				/*
1892 				 * otherwise we have one bad data stripe and
1893 				 * a good P stripe.  raid5!
1894 				 */
1895 				goto pstripe;
1896 			}
1897 
1898 			if (rbio->bbio->raid_map[failb] == RAID5_P_STRIPE) {
1899 				raid6_datap_recov(rbio->real_stripes,
1900 						  PAGE_SIZE, faila, pointers);
1901 			} else {
1902 				raid6_2data_recov(rbio->real_stripes,
1903 						  PAGE_SIZE, faila, failb,
1904 						  pointers);
1905 			}
1906 		} else {
1907 			void *p;
1908 
1909 			/* rebuild from P stripe here (raid5 or raid6) */
1910 			BUG_ON(failb != -1);
1911 pstripe:
1912 			/* Copy parity block into failed block to start with */
1913 			copy_page(pointers[faila], pointers[rbio->nr_data]);
1914 
1915 			/* rearrange the pointer array */
1916 			p = pointers[faila];
1917 			for (stripe = faila; stripe < rbio->nr_data - 1; stripe++)
1918 				pointers[stripe] = pointers[stripe + 1];
1919 			pointers[rbio->nr_data - 1] = p;
1920 
1921 			/* xor in the rest */
1922 			run_xor(pointers, rbio->nr_data - 1, PAGE_SIZE);
1923 		}
1924 		/* if we're doing this rebuild as part of an rmw, go through
1925 		 * and set all of our private rbio pages in the
1926 		 * failed stripes as uptodate.  This way finish_rmw will
1927 		 * know they can be trusted.  If this was a read reconstruction,
1928 		 * other endio functions will fiddle the uptodate bits
1929 		 */
1930 		if (rbio->operation == BTRFS_RBIO_WRITE) {
1931 			for (i = 0;  i < rbio->stripe_npages; i++) {
1932 				if (faila != -1) {
1933 					page = rbio_stripe_page(rbio, faila, i);
1934 					SetPageUptodate(page);
1935 				}
1936 				if (failb != -1) {
1937 					page = rbio_stripe_page(rbio, failb, i);
1938 					SetPageUptodate(page);
1939 				}
1940 			}
1941 		}
1942 		for (stripe = 0; stripe < rbio->real_stripes; stripe++) {
1943 			/*
1944 			 * if we're rebuilding a read, we have to use
1945 			 * pages from the bio list
1946 			 */
1947 			if ((rbio->operation == BTRFS_RBIO_READ_REBUILD ||
1948 			     rbio->operation == BTRFS_RBIO_REBUILD_MISSING) &&
1949 			    (stripe == faila || stripe == failb)) {
1950 				page = page_in_rbio(rbio, stripe, pagenr, 0);
1951 			} else {
1952 				page = rbio_stripe_page(rbio, stripe, pagenr);
1953 			}
1954 			kunmap(page);
1955 		}
1956 	}
1957 
1958 	err = BLK_STS_OK;
1959 cleanup:
1960 	kfree(pointers);
1961 
1962 cleanup_io:
1963 	/*
1964 	 * Similar to READ_REBUILD, REBUILD_MISSING at this point also has a
1965 	 * valid rbio which is consistent with ondisk content, thus such a
1966 	 * valid rbio can be cached to avoid further disk reads.
1967 	 */
1968 	if (rbio->operation == BTRFS_RBIO_READ_REBUILD ||
1969 	    rbio->operation == BTRFS_RBIO_REBUILD_MISSING) {
1970 		/*
1971 		 * - In case of two failures, where rbio->failb != -1:
1972 		 *
1973 		 *   Do not cache this rbio since the above read reconstruction
1974 		 *   (raid6_datap_recov() or raid6_2data_recov()) may have
1975 		 *   changed some content of stripes which are not identical to
1976 		 *   on-disk content any more, otherwise, a later write/recover
1977 		 *   may steal stripe_pages from this rbio and end up with
1978 		 *   corruptions or rebuild failures.
1979 		 *
1980 		 * - In case of single failure, where rbio->failb == -1:
1981 		 *
1982 		 *   Cache this rbio iff the above read reconstruction is
1983 		 *   excuted without problems.
1984 		 */
1985 		if (err == BLK_STS_OK && rbio->failb < 0)
1986 			cache_rbio_pages(rbio);
1987 		else
1988 			clear_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
1989 
1990 		rbio_orig_end_io(rbio, err);
1991 	} else if (err == BLK_STS_OK) {
1992 		rbio->faila = -1;
1993 		rbio->failb = -1;
1994 
1995 		if (rbio->operation == BTRFS_RBIO_WRITE)
1996 			finish_rmw(rbio);
1997 		else if (rbio->operation == BTRFS_RBIO_PARITY_SCRUB)
1998 			finish_parity_scrub(rbio, 0);
1999 		else
2000 			BUG();
2001 	} else {
2002 		rbio_orig_end_io(rbio, err);
2003 	}
2004 }
2005 
2006 /*
2007  * This is called only for stripes we've read from disk to
2008  * reconstruct the parity.
2009  */
2010 static void raid_recover_end_io(struct bio *bio)
2011 {
2012 	struct btrfs_raid_bio *rbio = bio->bi_private;
2013 
2014 	/*
2015 	 * we only read stripe pages off the disk, set them
2016 	 * up to date if there were no errors
2017 	 */
2018 	if (bio->bi_status)
2019 		fail_bio_stripe(rbio, bio);
2020 	else
2021 		set_bio_pages_uptodate(bio);
2022 	bio_put(bio);
2023 
2024 	if (!atomic_dec_and_test(&rbio->stripes_pending))
2025 		return;
2026 
2027 	if (atomic_read(&rbio->error) > rbio->bbio->max_errors)
2028 		rbio_orig_end_io(rbio, BLK_STS_IOERR);
2029 	else
2030 		__raid_recover_end_io(rbio);
2031 }
2032 
2033 /*
2034  * reads everything we need off the disk to reconstruct
2035  * the parity. endio handlers trigger final reconstruction
2036  * when the IO is done.
2037  *
2038  * This is used both for reads from the higher layers and for
2039  * parity construction required to finish a rmw cycle.
2040  */
2041 static int __raid56_parity_recover(struct btrfs_raid_bio *rbio)
2042 {
2043 	int bios_to_read = 0;
2044 	struct bio_list bio_list;
2045 	int ret;
2046 	int pagenr;
2047 	int stripe;
2048 	struct bio *bio;
2049 
2050 	bio_list_init(&bio_list);
2051 
2052 	ret = alloc_rbio_pages(rbio);
2053 	if (ret)
2054 		goto cleanup;
2055 
2056 	atomic_set(&rbio->error, 0);
2057 
2058 	/*
2059 	 * read everything that hasn't failed.  Thanks to the
2060 	 * stripe cache, it is possible that some or all of these
2061 	 * pages are going to be uptodate.
2062 	 */
2063 	for (stripe = 0; stripe < rbio->real_stripes; stripe++) {
2064 		if (rbio->faila == stripe || rbio->failb == stripe) {
2065 			atomic_inc(&rbio->error);
2066 			continue;
2067 		}
2068 
2069 		for (pagenr = 0; pagenr < rbio->stripe_npages; pagenr++) {
2070 			struct page *p;
2071 
2072 			/*
2073 			 * the rmw code may have already read this
2074 			 * page in
2075 			 */
2076 			p = rbio_stripe_page(rbio, stripe, pagenr);
2077 			if (PageUptodate(p))
2078 				continue;
2079 
2080 			ret = rbio_add_io_page(rbio, &bio_list,
2081 				       rbio_stripe_page(rbio, stripe, pagenr),
2082 				       stripe, pagenr, rbio->stripe_len);
2083 			if (ret < 0)
2084 				goto cleanup;
2085 		}
2086 	}
2087 
2088 	bios_to_read = bio_list_size(&bio_list);
2089 	if (!bios_to_read) {
2090 		/*
2091 		 * we might have no bios to read just because the pages
2092 		 * were up to date, or we might have no bios to read because
2093 		 * the devices were gone.
2094 		 */
2095 		if (atomic_read(&rbio->error) <= rbio->bbio->max_errors) {
2096 			__raid_recover_end_io(rbio);
2097 			goto out;
2098 		} else {
2099 			goto cleanup;
2100 		}
2101 	}
2102 
2103 	/*
2104 	 * the bbio may be freed once we submit the last bio.  Make sure
2105 	 * not to touch it after that
2106 	 */
2107 	atomic_set(&rbio->stripes_pending, bios_to_read);
2108 	while (1) {
2109 		bio = bio_list_pop(&bio_list);
2110 		if (!bio)
2111 			break;
2112 
2113 		bio->bi_private = rbio;
2114 		bio->bi_end_io = raid_recover_end_io;
2115 		bio->bi_opf = REQ_OP_READ;
2116 
2117 		btrfs_bio_wq_end_io(rbio->fs_info, bio, BTRFS_WQ_ENDIO_RAID56);
2118 
2119 		submit_bio(bio);
2120 	}
2121 out:
2122 	return 0;
2123 
2124 cleanup:
2125 	if (rbio->operation == BTRFS_RBIO_READ_REBUILD ||
2126 	    rbio->operation == BTRFS_RBIO_REBUILD_MISSING)
2127 		rbio_orig_end_io(rbio, BLK_STS_IOERR);
2128 
2129 	while ((bio = bio_list_pop(&bio_list)))
2130 		bio_put(bio);
2131 
2132 	return -EIO;
2133 }
2134 
2135 /*
2136  * the main entry point for reads from the higher layers.  This
2137  * is really only called when the normal read path had a failure,
2138  * so we assume the bio they send down corresponds to a failed part
2139  * of the drive.
2140  */
2141 int raid56_parity_recover(struct btrfs_fs_info *fs_info, struct bio *bio,
2142 			  struct btrfs_bio *bbio, u64 stripe_len,
2143 			  int mirror_num, int generic_io)
2144 {
2145 	struct btrfs_raid_bio *rbio;
2146 	int ret;
2147 
2148 	if (generic_io) {
2149 		ASSERT(bbio->mirror_num == mirror_num);
2150 		btrfs_io_bio(bio)->mirror_num = mirror_num;
2151 	}
2152 
2153 	rbio = alloc_rbio(fs_info, bbio, stripe_len);
2154 	if (IS_ERR(rbio)) {
2155 		if (generic_io)
2156 			btrfs_put_bbio(bbio);
2157 		return PTR_ERR(rbio);
2158 	}
2159 
2160 	rbio->operation = BTRFS_RBIO_READ_REBUILD;
2161 	bio_list_add(&rbio->bio_list, bio);
2162 	rbio->bio_list_bytes = bio->bi_iter.bi_size;
2163 
2164 	rbio->faila = find_logical_bio_stripe(rbio, bio);
2165 	if (rbio->faila == -1) {
2166 		btrfs_warn(fs_info,
2167 	"%s could not find the bad stripe in raid56 so that we cannot recover any more (bio has logical %llu len %llu, bbio has map_type %llu)",
2168 			   __func__, (u64)bio->bi_iter.bi_sector << 9,
2169 			   (u64)bio->bi_iter.bi_size, bbio->map_type);
2170 		if (generic_io)
2171 			btrfs_put_bbio(bbio);
2172 		kfree(rbio);
2173 		return -EIO;
2174 	}
2175 
2176 	if (generic_io) {
2177 		btrfs_bio_counter_inc_noblocked(fs_info);
2178 		rbio->generic_bio_cnt = 1;
2179 	} else {
2180 		btrfs_get_bbio(bbio);
2181 	}
2182 
2183 	/*
2184 	 * Loop retry:
2185 	 * for 'mirror == 2', reconstruct from all other stripes.
2186 	 * for 'mirror_num > 2', select a stripe to fail on every retry.
2187 	 */
2188 	if (mirror_num > 2) {
2189 		/*
2190 		 * 'mirror == 3' is to fail the p stripe and
2191 		 * reconstruct from the q stripe.  'mirror > 3' is to
2192 		 * fail a data stripe and reconstruct from p+q stripe.
2193 		 */
2194 		rbio->failb = rbio->real_stripes - (mirror_num - 1);
2195 		ASSERT(rbio->failb > 0);
2196 		if (rbio->failb <= rbio->faila)
2197 			rbio->failb--;
2198 	}
2199 
2200 	ret = lock_stripe_add(rbio);
2201 
2202 	/*
2203 	 * __raid56_parity_recover will end the bio with
2204 	 * any errors it hits.  We don't want to return
2205 	 * its error value up the stack because our caller
2206 	 * will end up calling bio_endio with any nonzero
2207 	 * return
2208 	 */
2209 	if (ret == 0)
2210 		__raid56_parity_recover(rbio);
2211 	/*
2212 	 * our rbio has been added to the list of
2213 	 * rbios that will be handled after the
2214 	 * currently lock owner is done
2215 	 */
2216 	return 0;
2217 
2218 }
2219 
2220 static void rmw_work(struct btrfs_work *work)
2221 {
2222 	struct btrfs_raid_bio *rbio;
2223 
2224 	rbio = container_of(work, struct btrfs_raid_bio, work);
2225 	raid56_rmw_stripe(rbio);
2226 }
2227 
2228 static void read_rebuild_work(struct btrfs_work *work)
2229 {
2230 	struct btrfs_raid_bio *rbio;
2231 
2232 	rbio = container_of(work, struct btrfs_raid_bio, work);
2233 	__raid56_parity_recover(rbio);
2234 }
2235 
2236 /*
2237  * The following code is used to scrub/replace the parity stripe
2238  *
2239  * Caller must have already increased bio_counter for getting @bbio.
2240  *
2241  * Note: We need make sure all the pages that add into the scrub/replace
2242  * raid bio are correct and not be changed during the scrub/replace. That
2243  * is those pages just hold metadata or file data with checksum.
2244  */
2245 
2246 struct btrfs_raid_bio *
2247 raid56_parity_alloc_scrub_rbio(struct btrfs_fs_info *fs_info, struct bio *bio,
2248 			       struct btrfs_bio *bbio, u64 stripe_len,
2249 			       struct btrfs_device *scrub_dev,
2250 			       unsigned long *dbitmap, int stripe_nsectors)
2251 {
2252 	struct btrfs_raid_bio *rbio;
2253 	int i;
2254 
2255 	rbio = alloc_rbio(fs_info, bbio, stripe_len);
2256 	if (IS_ERR(rbio))
2257 		return NULL;
2258 	bio_list_add(&rbio->bio_list, bio);
2259 	/*
2260 	 * This is a special bio which is used to hold the completion handler
2261 	 * and make the scrub rbio is similar to the other types
2262 	 */
2263 	ASSERT(!bio->bi_iter.bi_size);
2264 	rbio->operation = BTRFS_RBIO_PARITY_SCRUB;
2265 
2266 	/*
2267 	 * After mapping bbio with BTRFS_MAP_WRITE, parities have been sorted
2268 	 * to the end position, so this search can start from the first parity
2269 	 * stripe.
2270 	 */
2271 	for (i = rbio->nr_data; i < rbio->real_stripes; i++) {
2272 		if (bbio->stripes[i].dev == scrub_dev) {
2273 			rbio->scrubp = i;
2274 			break;
2275 		}
2276 	}
2277 	ASSERT(i < rbio->real_stripes);
2278 
2279 	/* Now we just support the sectorsize equals to page size */
2280 	ASSERT(fs_info->sectorsize == PAGE_SIZE);
2281 	ASSERT(rbio->stripe_npages == stripe_nsectors);
2282 	bitmap_copy(rbio->dbitmap, dbitmap, stripe_nsectors);
2283 
2284 	/*
2285 	 * We have already increased bio_counter when getting bbio, record it
2286 	 * so we can free it at rbio_orig_end_io().
2287 	 */
2288 	rbio->generic_bio_cnt = 1;
2289 
2290 	return rbio;
2291 }
2292 
2293 /* Used for both parity scrub and missing. */
2294 void raid56_add_scrub_pages(struct btrfs_raid_bio *rbio, struct page *page,
2295 			    u64 logical)
2296 {
2297 	int stripe_offset;
2298 	int index;
2299 
2300 	ASSERT(logical >= rbio->bbio->raid_map[0]);
2301 	ASSERT(logical + PAGE_SIZE <= rbio->bbio->raid_map[0] +
2302 				rbio->stripe_len * rbio->nr_data);
2303 	stripe_offset = (int)(logical - rbio->bbio->raid_map[0]);
2304 	index = stripe_offset >> PAGE_SHIFT;
2305 	rbio->bio_pages[index] = page;
2306 }
2307 
2308 /*
2309  * We just scrub the parity that we have correct data on the same horizontal,
2310  * so we needn't allocate all pages for all the stripes.
2311  */
2312 static int alloc_rbio_essential_pages(struct btrfs_raid_bio *rbio)
2313 {
2314 	int i;
2315 	int bit;
2316 	int index;
2317 	struct page *page;
2318 
2319 	for_each_set_bit(bit, rbio->dbitmap, rbio->stripe_npages) {
2320 		for (i = 0; i < rbio->real_stripes; i++) {
2321 			index = i * rbio->stripe_npages + bit;
2322 			if (rbio->stripe_pages[index])
2323 				continue;
2324 
2325 			page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
2326 			if (!page)
2327 				return -ENOMEM;
2328 			rbio->stripe_pages[index] = page;
2329 		}
2330 	}
2331 	return 0;
2332 }
2333 
2334 static noinline void finish_parity_scrub(struct btrfs_raid_bio *rbio,
2335 					 int need_check)
2336 {
2337 	struct btrfs_bio *bbio = rbio->bbio;
2338 	void **pointers = rbio->finish_pointers;
2339 	unsigned long *pbitmap = rbio->finish_pbitmap;
2340 	int nr_data = rbio->nr_data;
2341 	int stripe;
2342 	int pagenr;
2343 	int p_stripe = -1;
2344 	int q_stripe = -1;
2345 	struct page *p_page = NULL;
2346 	struct page *q_page = NULL;
2347 	struct bio_list bio_list;
2348 	struct bio *bio;
2349 	int is_replace = 0;
2350 	int ret;
2351 
2352 	bio_list_init(&bio_list);
2353 
2354 	if (rbio->real_stripes - rbio->nr_data == 1) {
2355 		p_stripe = rbio->real_stripes - 1;
2356 	} else if (rbio->real_stripes - rbio->nr_data == 2) {
2357 		p_stripe = rbio->real_stripes - 2;
2358 		q_stripe = rbio->real_stripes - 1;
2359 	} else {
2360 		BUG();
2361 	}
2362 
2363 	if (bbio->num_tgtdevs && bbio->tgtdev_map[rbio->scrubp]) {
2364 		is_replace = 1;
2365 		bitmap_copy(pbitmap, rbio->dbitmap, rbio->stripe_npages);
2366 	}
2367 
2368 	/*
2369 	 * Because the higher layers(scrubber) are unlikely to
2370 	 * use this area of the disk again soon, so don't cache
2371 	 * it.
2372 	 */
2373 	clear_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
2374 
2375 	if (!need_check)
2376 		goto writeback;
2377 
2378 	p_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
2379 	if (!p_page)
2380 		goto cleanup;
2381 	SetPageUptodate(p_page);
2382 
2383 	if (q_stripe != -1) {
2384 		q_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
2385 		if (!q_page) {
2386 			__free_page(p_page);
2387 			goto cleanup;
2388 		}
2389 		SetPageUptodate(q_page);
2390 	}
2391 
2392 	atomic_set(&rbio->error, 0);
2393 
2394 	for_each_set_bit(pagenr, rbio->dbitmap, rbio->stripe_npages) {
2395 		struct page *p;
2396 		void *parity;
2397 		/* first collect one page from each data stripe */
2398 		for (stripe = 0; stripe < nr_data; stripe++) {
2399 			p = page_in_rbio(rbio, stripe, pagenr, 0);
2400 			pointers[stripe] = kmap(p);
2401 		}
2402 
2403 		/* then add the parity stripe */
2404 		pointers[stripe++] = kmap(p_page);
2405 
2406 		if (q_stripe != -1) {
2407 
2408 			/*
2409 			 * raid6, add the qstripe and call the
2410 			 * library function to fill in our p/q
2411 			 */
2412 			pointers[stripe++] = kmap(q_page);
2413 
2414 			raid6_call.gen_syndrome(rbio->real_stripes, PAGE_SIZE,
2415 						pointers);
2416 		} else {
2417 			/* raid5 */
2418 			copy_page(pointers[nr_data], pointers[0]);
2419 			run_xor(pointers + 1, nr_data - 1, PAGE_SIZE);
2420 		}
2421 
2422 		/* Check scrubbing parity and repair it */
2423 		p = rbio_stripe_page(rbio, rbio->scrubp, pagenr);
2424 		parity = kmap(p);
2425 		if (memcmp(parity, pointers[rbio->scrubp], PAGE_SIZE))
2426 			copy_page(parity, pointers[rbio->scrubp]);
2427 		else
2428 			/* Parity is right, needn't writeback */
2429 			bitmap_clear(rbio->dbitmap, pagenr, 1);
2430 		kunmap(p);
2431 
2432 		for (stripe = 0; stripe < rbio->real_stripes; stripe++)
2433 			kunmap(page_in_rbio(rbio, stripe, pagenr, 0));
2434 	}
2435 
2436 	__free_page(p_page);
2437 	if (q_page)
2438 		__free_page(q_page);
2439 
2440 writeback:
2441 	/*
2442 	 * time to start writing.  Make bios for everything from the
2443 	 * higher layers (the bio_list in our rbio) and our p/q.  Ignore
2444 	 * everything else.
2445 	 */
2446 	for_each_set_bit(pagenr, rbio->dbitmap, rbio->stripe_npages) {
2447 		struct page *page;
2448 
2449 		page = rbio_stripe_page(rbio, rbio->scrubp, pagenr);
2450 		ret = rbio_add_io_page(rbio, &bio_list,
2451 			       page, rbio->scrubp, pagenr, rbio->stripe_len);
2452 		if (ret)
2453 			goto cleanup;
2454 	}
2455 
2456 	if (!is_replace)
2457 		goto submit_write;
2458 
2459 	for_each_set_bit(pagenr, pbitmap, rbio->stripe_npages) {
2460 		struct page *page;
2461 
2462 		page = rbio_stripe_page(rbio, rbio->scrubp, pagenr);
2463 		ret = rbio_add_io_page(rbio, &bio_list, page,
2464 				       bbio->tgtdev_map[rbio->scrubp],
2465 				       pagenr, rbio->stripe_len);
2466 		if (ret)
2467 			goto cleanup;
2468 	}
2469 
2470 submit_write:
2471 	nr_data = bio_list_size(&bio_list);
2472 	if (!nr_data) {
2473 		/* Every parity is right */
2474 		rbio_orig_end_io(rbio, BLK_STS_OK);
2475 		return;
2476 	}
2477 
2478 	atomic_set(&rbio->stripes_pending, nr_data);
2479 
2480 	while (1) {
2481 		bio = bio_list_pop(&bio_list);
2482 		if (!bio)
2483 			break;
2484 
2485 		bio->bi_private = rbio;
2486 		bio->bi_end_io = raid_write_end_io;
2487 		bio->bi_opf = REQ_OP_WRITE;
2488 
2489 		submit_bio(bio);
2490 	}
2491 	return;
2492 
2493 cleanup:
2494 	rbio_orig_end_io(rbio, BLK_STS_IOERR);
2495 
2496 	while ((bio = bio_list_pop(&bio_list)))
2497 		bio_put(bio);
2498 }
2499 
2500 static inline int is_data_stripe(struct btrfs_raid_bio *rbio, int stripe)
2501 {
2502 	if (stripe >= 0 && stripe < rbio->nr_data)
2503 		return 1;
2504 	return 0;
2505 }
2506 
2507 /*
2508  * While we're doing the parity check and repair, we could have errors
2509  * in reading pages off the disk.  This checks for errors and if we're
2510  * not able to read the page it'll trigger parity reconstruction.  The
2511  * parity scrub will be finished after we've reconstructed the failed
2512  * stripes
2513  */
2514 static void validate_rbio_for_parity_scrub(struct btrfs_raid_bio *rbio)
2515 {
2516 	if (atomic_read(&rbio->error) > rbio->bbio->max_errors)
2517 		goto cleanup;
2518 
2519 	if (rbio->faila >= 0 || rbio->failb >= 0) {
2520 		int dfail = 0, failp = -1;
2521 
2522 		if (is_data_stripe(rbio, rbio->faila))
2523 			dfail++;
2524 		else if (is_parity_stripe(rbio->faila))
2525 			failp = rbio->faila;
2526 
2527 		if (is_data_stripe(rbio, rbio->failb))
2528 			dfail++;
2529 		else if (is_parity_stripe(rbio->failb))
2530 			failp = rbio->failb;
2531 
2532 		/*
2533 		 * Because we can not use a scrubbing parity to repair
2534 		 * the data, so the capability of the repair is declined.
2535 		 * (In the case of RAID5, we can not repair anything)
2536 		 */
2537 		if (dfail > rbio->bbio->max_errors - 1)
2538 			goto cleanup;
2539 
2540 		/*
2541 		 * If all data is good, only parity is correctly, just
2542 		 * repair the parity.
2543 		 */
2544 		if (dfail == 0) {
2545 			finish_parity_scrub(rbio, 0);
2546 			return;
2547 		}
2548 
2549 		/*
2550 		 * Here means we got one corrupted data stripe and one
2551 		 * corrupted parity on RAID6, if the corrupted parity
2552 		 * is scrubbing parity, luckily, use the other one to repair
2553 		 * the data, or we can not repair the data stripe.
2554 		 */
2555 		if (failp != rbio->scrubp)
2556 			goto cleanup;
2557 
2558 		__raid_recover_end_io(rbio);
2559 	} else {
2560 		finish_parity_scrub(rbio, 1);
2561 	}
2562 	return;
2563 
2564 cleanup:
2565 	rbio_orig_end_io(rbio, BLK_STS_IOERR);
2566 }
2567 
2568 /*
2569  * end io for the read phase of the rmw cycle.  All the bios here are physical
2570  * stripe bios we've read from the disk so we can recalculate the parity of the
2571  * stripe.
2572  *
2573  * This will usually kick off finish_rmw once all the bios are read in, but it
2574  * may trigger parity reconstruction if we had any errors along the way
2575  */
2576 static void raid56_parity_scrub_end_io(struct bio *bio)
2577 {
2578 	struct btrfs_raid_bio *rbio = bio->bi_private;
2579 
2580 	if (bio->bi_status)
2581 		fail_bio_stripe(rbio, bio);
2582 	else
2583 		set_bio_pages_uptodate(bio);
2584 
2585 	bio_put(bio);
2586 
2587 	if (!atomic_dec_and_test(&rbio->stripes_pending))
2588 		return;
2589 
2590 	/*
2591 	 * this will normally call finish_rmw to start our write
2592 	 * but if there are any failed stripes we'll reconstruct
2593 	 * from parity first
2594 	 */
2595 	validate_rbio_for_parity_scrub(rbio);
2596 }
2597 
2598 static void raid56_parity_scrub_stripe(struct btrfs_raid_bio *rbio)
2599 {
2600 	int bios_to_read = 0;
2601 	struct bio_list bio_list;
2602 	int ret;
2603 	int pagenr;
2604 	int stripe;
2605 	struct bio *bio;
2606 
2607 	bio_list_init(&bio_list);
2608 
2609 	ret = alloc_rbio_essential_pages(rbio);
2610 	if (ret)
2611 		goto cleanup;
2612 
2613 	atomic_set(&rbio->error, 0);
2614 	/*
2615 	 * build a list of bios to read all the missing parts of this
2616 	 * stripe
2617 	 */
2618 	for (stripe = 0; stripe < rbio->real_stripes; stripe++) {
2619 		for_each_set_bit(pagenr, rbio->dbitmap, rbio->stripe_npages) {
2620 			struct page *page;
2621 			/*
2622 			 * we want to find all the pages missing from
2623 			 * the rbio and read them from the disk.  If
2624 			 * page_in_rbio finds a page in the bio list
2625 			 * we don't need to read it off the stripe.
2626 			 */
2627 			page = page_in_rbio(rbio, stripe, pagenr, 1);
2628 			if (page)
2629 				continue;
2630 
2631 			page = rbio_stripe_page(rbio, stripe, pagenr);
2632 			/*
2633 			 * the bio cache may have handed us an uptodate
2634 			 * page.  If so, be happy and use it
2635 			 */
2636 			if (PageUptodate(page))
2637 				continue;
2638 
2639 			ret = rbio_add_io_page(rbio, &bio_list, page,
2640 				       stripe, pagenr, rbio->stripe_len);
2641 			if (ret)
2642 				goto cleanup;
2643 		}
2644 	}
2645 
2646 	bios_to_read = bio_list_size(&bio_list);
2647 	if (!bios_to_read) {
2648 		/*
2649 		 * this can happen if others have merged with
2650 		 * us, it means there is nothing left to read.
2651 		 * But if there are missing devices it may not be
2652 		 * safe to do the full stripe write yet.
2653 		 */
2654 		goto finish;
2655 	}
2656 
2657 	/*
2658 	 * the bbio may be freed once we submit the last bio.  Make sure
2659 	 * not to touch it after that
2660 	 */
2661 	atomic_set(&rbio->stripes_pending, bios_to_read);
2662 	while (1) {
2663 		bio = bio_list_pop(&bio_list);
2664 		if (!bio)
2665 			break;
2666 
2667 		bio->bi_private = rbio;
2668 		bio->bi_end_io = raid56_parity_scrub_end_io;
2669 		bio->bi_opf = REQ_OP_READ;
2670 
2671 		btrfs_bio_wq_end_io(rbio->fs_info, bio, BTRFS_WQ_ENDIO_RAID56);
2672 
2673 		submit_bio(bio);
2674 	}
2675 	/* the actual write will happen once the reads are done */
2676 	return;
2677 
2678 cleanup:
2679 	rbio_orig_end_io(rbio, BLK_STS_IOERR);
2680 
2681 	while ((bio = bio_list_pop(&bio_list)))
2682 		bio_put(bio);
2683 
2684 	return;
2685 
2686 finish:
2687 	validate_rbio_for_parity_scrub(rbio);
2688 }
2689 
2690 static void scrub_parity_work(struct btrfs_work *work)
2691 {
2692 	struct btrfs_raid_bio *rbio;
2693 
2694 	rbio = container_of(work, struct btrfs_raid_bio, work);
2695 	raid56_parity_scrub_stripe(rbio);
2696 }
2697 
2698 void raid56_parity_submit_scrub_rbio(struct btrfs_raid_bio *rbio)
2699 {
2700 	if (!lock_stripe_add(rbio))
2701 		start_async_work(rbio, scrub_parity_work);
2702 }
2703 
2704 /* The following code is used for dev replace of a missing RAID 5/6 device. */
2705 
2706 struct btrfs_raid_bio *
2707 raid56_alloc_missing_rbio(struct btrfs_fs_info *fs_info, struct bio *bio,
2708 			  struct btrfs_bio *bbio, u64 length)
2709 {
2710 	struct btrfs_raid_bio *rbio;
2711 
2712 	rbio = alloc_rbio(fs_info, bbio, length);
2713 	if (IS_ERR(rbio))
2714 		return NULL;
2715 
2716 	rbio->operation = BTRFS_RBIO_REBUILD_MISSING;
2717 	bio_list_add(&rbio->bio_list, bio);
2718 	/*
2719 	 * This is a special bio which is used to hold the completion handler
2720 	 * and make the scrub rbio is similar to the other types
2721 	 */
2722 	ASSERT(!bio->bi_iter.bi_size);
2723 
2724 	rbio->faila = find_logical_bio_stripe(rbio, bio);
2725 	if (rbio->faila == -1) {
2726 		BUG();
2727 		kfree(rbio);
2728 		return NULL;
2729 	}
2730 
2731 	/*
2732 	 * When we get bbio, we have already increased bio_counter, record it
2733 	 * so we can free it at rbio_orig_end_io()
2734 	 */
2735 	rbio->generic_bio_cnt = 1;
2736 
2737 	return rbio;
2738 }
2739 
2740 void raid56_submit_missing_rbio(struct btrfs_raid_bio *rbio)
2741 {
2742 	if (!lock_stripe_add(rbio))
2743 		start_async_work(rbio, read_rebuild_work);
2744 }
2745